Conversational AI and NLP

Sharing session at Singapore Tourism Board - April 2019

Transcript

1. Natural Language Processing and Conversational AI David Low Co-founder / CDS
2. Bio • Research Urban Mobility | Social Media • Public Service GovTech(IDA) Data Science Division • Teach Deep Learning Masterclass at National University of Singapore (NUS) • Startup Conversational AI | Deep Natural Language Processing
3. Rising trend of mobile (& messaging) usage
4. As messaging apps have become indispensable parts of our lives, businesses are determined to be where their users are.
6. Source: Apartment Ocean A chatbot
7. What is Natural Language Processing?
8. What is Natural Language Processing? • is a field at the intersection of Computer Science, Artificial Intelligence (AI) and Linguistics. • Goal For computers to process or “understand” natural language in order to perform tasks that are useful • Not to be confused with “Computational Linguistics” • Deep NLP = Deep Learning based Natural Language Processing
9. Example I - Sentiment Analysis
10. Example II - Machine Translation Source: Google AI
11. NLU (Understanding) vs NLP (Processing) Source: Stanford NLP Group
12. Current State of Conversational AI
13. “While Siri, Alexa and the likes can follow simple spoken or typed commands and answer basic questions; they can’t hold a conversation and have no real understanding of the words they use. ” Will Knight
14. In the past... Chatbot A Chatbot B
15. Now Source: Salemove
16. Why is NLP difficult? “Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo” - This is a grammatically correct sentence in American English.
17. What is Homonym? (In linguistics) Words with identical pronunciation and spelling, whilst maintaining different meanings.
18. Buffalo - a homonym 1 - Noun: The city of Buffalo, at western New York state, US. 2 - Noun: The animal, Buffalo. 3 - Verb: To confuse, deceive or intimidate.
19. Three groups of buffalo “Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo”
20. Why is NLP difficult? [cont] • Sheer complexity of sentence structure • Ambiguity ◦ Eg: “I made her duck” • Meaning is context sensitive ◦ Depends on the people present e.g. “How far is it?” (miles, km?) ◦ Depends on the time of day, e.g. “Let's go eat” ◦ Depends on prior sentences: “The third one” • Recognizing named entities (people/places/…) • Slang, jargon, humour, sarcasm, spelling mistakes, grammar mistakes and abbreviations… Source: Stanford CS224
21. Recent Developments
22. Transfer Learning • Significant breakthroughs in NLP field in 2018 (BERT, ELMO, ULMFiT etc)
23. In near future
24. Deal or no deal? Training AI bots to negotiate Source: FAIR
25. Dialog agents with ability to negotiate
26. Microsoft’s Tay: How chatbot goes bad • Experiment on conversational understanding. • Fully automated self-learning • Garbage in, garbage out! In less than 24 hours
27. What is Conversational UX? Conversational UX is the way a user experiences an interaction with a business or a service through an automated conversational interface that learns along the way.
29. Good or Bad Conversational UX? User: “I need blue shoes” Chatbot: “Great. Please use the command ITEM - COLOUR - SIZE - BRAND - TYPE” User: “Ok I want blue Nike shoes size 8 for sports.” Chatbot: “Sorry, I did not get that. Please use the command ITEM - COLOUR - SIZE - BRAND - TYPE” User: “Shoes blue 8 Nike sports” Chatbot: “Sorry, I did not get that. Please use the command ITEM - COLOUR - SIZE - BRAND - TYPE” User: “SHOES - BLUE - 8 - NIKE - SPORT” Chatbot: “Here is a selection for you to pick from...”
30. “Good conversational UX should be a seamless experience, that makes people feel like talking to another human without realizing he/she is talking to a bot”
31. https://www.youtube.com/watch?time_continue=7&v=znNe4pMCsD4
32. Google Duplex • Trained within a closed and narrow domain • Incorporation of speech disfluencies (e.g. “hmm”s and “uh”s) to make it sounds more natural Source: Google